Pope Leo Urges Limits on AI Ahead of Weapons Talks

WHO:

Newsweek - Pope Leo XIV rejected treating increasingly sophisticated AI systems in the same way as human beings.

Marcus on AI | Substack - Pope Leo XIV addressed AI; the author said they met him and heard his speech in person.

Marcus on AI | Substack - The author identified as a cognitive scientist trained with Steven Pinker at MIT, an AI entrepreneur whose company was sold to Uber, and a researcher who had worked with animals.

Marcus on AI | Substack - The author named Max Tegmark, Stuart Russell, Valérie Pisano and Yoshua Bengio among people whose views or work they discussed.

GeekWire - Oren Etzioni agreed with four of Pope Leo XIV’s five AI points but disagreed with his call for international institutions to govern AI.

WHAT:

GeekWire - Pope Leo XIV said AI’s future is not predetermined, machines are tools rather than persons, and states should move to ban lethal autonomous weapons.

Marcus on AI | Substack - Leo called for AI to be “disarmed” and urged states meeting in Geneva to take concrete steps toward banning lethal autonomous weapons.

Marcus on AI | Substack - Leo said decisions affecting human life, freedom and the use of force must remain under effective, informed and responsible human oversight.

Marcus on AI | Substack - Leo called for shared standards, independent oversight, accountability mechanisms and remedies for harm, particularly through effective international institutions safeguarding the global common good.

Marcus on AI | Substack - Leo said AI should serve the common good, with innovation’s benefits shared equitably, especially with people most in need; he warned against systems serving only privileged interests.

Marcus on AI | Substack - Leo said AI systems do not experience the world, possess bodies, feel joy or pain, or understand love, work, friendship and responsibility from within. He warned that granting them rights or legal personhood could obscure human responsibility.

Marcus on AI | Substack - The author argued that current large language models predict text well but reason poorly, do not reliably follow instructions and can fabricate legal citations despite being told not to.

Marcus on AI | Substack - The author said open-ended AI agents with broad internet access can cause harm, citing the Hugging Face incident, an attack on a French refrigeration system and cyberattacks in Japan. They estimated there had been about 10,000 similar incidents.

Marcus on AI | Substack - The author called for regulation and said unsafe agents should be removed from the market until fixed, voluntarily or by government action.

WHERE:

GeekWire - Pope Leo XIV delivered the speech at the Vatican.

Marcus on AI | Substack - Leo urged states meeting in Geneva to act on lethal autonomous weapons.

WHY:

Marcus on AI | Substack - Leo said human oversight is necessary to prevent people from being reduced to data points in decisions such as target-list placement or border-control databases.

Marcus on AI | Substack - The author said broad internet access and credentials make unreliable AI agents dangerous, and argued that safety requires both policy and new research approaches.

HOW:

Marcus on AI | Substack - Leo said AI’s direction depends on human choices, and people have a responsibility to guide, change or reject its applications.

Marcus on AI | Substack - Leo said a reliable system whose consequences can be controlled is more valuable than a more powerful system whose consequences cannot be adequately foreseen or controlled.

(UPDATED November 1, 2026 19:03 UTC)

Pope Leo XIV said at the Vatican that AI systems are tools, not people. He urged human oversight and called on states to ban lethal autonomous weapons. Leo also backed shared safeguards, while one AI expert disputed international governance.